Computers Chem. Engng Vol. 22, No. 78, pp.

867877, 1998
1998 Elsevier Science Ltd
All rights reserved. Printed in Great Britain
PII: S0098-1354(98)00030-1 00981354/98 $19.00#0.00
An industrial design/control study
for the vinyl acetate monomer process
Michael L. Luyben* and Bjo rn D. Tyre us
DuPont Central Research & Development, Experimental Station - Bldg 357, P.O. Box 80357,
Wilmington, DE 19880-0357, USA
(Received 23 April 1997; received in revised form 29 December 1997)
Abstract
This work presents design details of an industrial process for the manufacture of vinyl acetate monomer. Our
purpose is to oer a realistic example that is uniquely suited for academic researchers pursuing simulation,
design, and control studies. The vinyl acetate process has common, real chemical components. It contains many
standard unit operations in a realistic owsheet. And it illustrates the types of systems of industrial research
interest in the areas of process design, optimization, simulation, and control. Vapor-phase reactions convert
ethylene, oxygen, and acetic acid into vinyl acetate with water and carbon dioxide as byproducts. The process
contains a packed tubular reactor, a feed-euent heat exchanger, an absorber, a vaporizer, an azeotropic
distillation column with decanter, and both gas and liquid recycle streams. All physical property, kinetic, and
owsheet data have been compiled from sources in the open literature. We detail the owsheet information
required to construct rigorous steady state and dynamic mathematical models of the process and present the
process control requirements and objectives. Finally, we briey describe the rigorous nonlinear dynamic
simulation we have constructed for this process using TMODS, DuPonts in-house dynamic simulator. Models
of this process have also been developed by Aspen Technology and Hyprotech in their commercial simulators
and are available directly from the vendors. 1998 Elsevier Science Ltd. All rights reserved
Keywords: industrial design/control study; vinyl acetate monomer process
1. Introduction
Downs and Vogel (1993) published an industrial
plantwide control test problem that has proved to be
a benecial service to the academic process control
community. A number of researchers have utilized
this example to test their ideas and technical develop-
ments. Many publications have appeared about the
Eastman process, which was provided as a dynamic
simulation in Fortran code. The problem has usefully
served as a realistic check on the industrial and practi-
cal relevance of the ever-increasing amount of process
control publications in chemical engineering.
We have heard of continued interest among aca-
demic researchers within the process design and con-
trol areas to have additional industrial examples of
realistic processes that can be used in assessing new
technology. We also recognize the need in the litera-
ture for plantwide design, optimization, and control
studies that grapple with: (1) a realistically large
*Corresponding author. E-mail: luybenml@esvax.dnet.
dupont.com.
process owsheet containing standard chemical unit
operations; (2) a process with the typical industrial
characteristics of recycle streams and energy integra-
tion; and (3) real nonideal chemical components.
This paper presents design details of an industrial
process for the production of vinyl acetate monomer
and thus goes a step beyond the Eastman process
control challenge problem. The reaction loop section
of the vinyl acetate process contains a owsheet and
unit operations that are typical of many chemical
plants. It has both gas and liquid recycle streams with
real components.
We have chosen to convey this study as if we had
been assigned the task of designing the control system
for a proposed new vinyl acetate process that is to be
built. We have been given a particular preliminary
design that has not been optimized. The design could
potentially be improved with modications to the
owsheet or design parameters.
The data we provide in this paper are what would
typically be available or easily obtainable: (1) kinetic
reaction parameters and physical property data, (2)
a owsheet structure with stream and equipment
information, and (3) the location of control valves
867
Fig. 1. Vinyl acetate monomer process owsheet.
included in the preliminary design. However, we
would not be given a linear transfer function model.
We certainly would not have a rigorous nonlinear
dynamic simulation. Steady state and dynamic mod-
els would have to be constructed by using a commer-
cial software simulation package or by writing the
code in some chosen programming language. As a re-
sult, unlike Downs and Vogel (1993), we do not make
available any code that simulates the vinyl acetate
process. Based upon the information provided in this
paper, models have been developed by Aspen Techno-
logy and Hyprotech in their commercial simulators.
We have ourselves built a rigorous nonlinear dy-
namic model of the process described in this paper
using TMODS, DuPonts in-house dynamic simula-
tor. This model has been used to verify the feasibility
of the simulation and to test the plantwide control
strategy proposed in Luyben et al. (1997). We believe
that this process should serve as a useful example for
researchers who are interested in working on indus-
trially relevant problems in simulation, design, and
control.
The industrial process for the vapor-phase manu-
facture of vinyl acetate monomer is quite common
(Daniels, 1989) and utilizes widely available raw ma-
terials. Vinyl acetate is used chiey as a monomer to
make polyvinyl acetate and other copolymers.
Hoechst-Celanese, Union Carbide, and Quantum
Chemical are reported US manufacturers. DuPont
also currently operates a vinyl acetate process at its
plant in LaPorte, Texas. To protect any proprietary
DuPont information, all of the physical property and
kinetic data, process owsheet information, and
modeling formulation in this work come from sources
in the open literature. We cite each source of data and
our process owsheet is based upon the description in
Report 15B by SRI International (1994). No relation,
either implied or intended, exists between this pub-
lished study and the DuPont process.
2. Vinyl acetate process
Figure 1 shows the eleven basic unit operations
proposed for the reaction section of the vinyl acetate
process, which is the focus of this study and the plant
we wish to design and operate. Three raw materials,
ethylene (C
`
H
"
), oxygen (O
`
), and acetic acid (HAc),
are converted into the vinyl acetate (VAc) product.
Water (H
`
O) and carbon dioxide (CO
`
) are byprod-
ucts. We assume that an inert component, ethane
(C
`
H
'
), enters with the fresh ethylene feed stream. We
consider the following two reactions:
C
`
H
"
#CH
`
COOH#1/2 O
`
PCH
`
"CHOCOCH
`
#H
`
O, (1)
C
`
H
"
#3O
`
P2CO
`
#2H
`
O. (2)
The exothermic reactions occur in a reactor contain-
ing tubes packed with a precious metal catalyst on
a silica support. Heat is removed from the reactor by
generating steam on the shell side of the tubes. Water
ows to the reactor from a steam drum, to which
make-up water (BFW) is supplied. The steam leaves
the drum as saturated vapor. The reactions are irre-
versible and the reaction rates have an Arrhenius-type
dependence on temperature.
868 M.L. LUYBEN and B.D. TYRE US
We located plots of experimental kinetic data in
Samanos et al. (1971) for a particular vinyl acetate
catalyst. As summarized by Neurock et al. (1996),
various mechanisms have been proposed for the
formation of vinyl acetate (i.e. Samanos et al., 1971;
and Nakamura and Yasui, 1970, develop completely
dierent expressions). However, we derived the fol-
lowing rate expressions that provide the best t to the
experimental data.
r

"0.1036 exp(!3674/)
;
p
-
p
#
p

(1#1.7p
5
)
(1#0.583p
-
(1#1.7p
5
)) (1#6.8p

)
, (3)
r
`
"1.9365;10` exp(!10, 116/)
;
p
-
(1#0.68p
5
)
1#0.76p
-
(1#0.68p
5
)
, (4)
where r

has units of moles of vinyl acetate produced/

min/(g catalyst) and r
`
has units of moles of ethylene
consumed/min/(g catalyst). is the absolute temper-
ature in Kelvin and p
G
is the partial pressure of com-
ponent i (O is oxygen, E is ethylene, A is acetic acid,
and is water) in psia.
The standard state heat of reaction is
!42.1 kcal/mol of vinyl acetate for r

and !316
kcal/mol of ethylene for r
`
. These values are cal-
culated using heats of formation from the DIPPR
database. Thus the reactions are quite exothermic,
particularly the combustion reaction to carbon dioxi-
de, which also is more sensitive to temperature be-
cause of the higher activation energy.
The reactor euent ows through a process-to-
process heat exchanger, where the cold stream is the
gas recycle. The reactor euent is then cooled with
cooling water and the vapor (oxygen, ethylene, carbon
dioxide, ethane) and liquid (vinyl acetate, water, acetic
acid) are separated. The vapor stream from the separ-
ator goes to the compressor and the liquid stream
from the separator becomes a part of the feed to the
azeotropic distillation column. The gas from the com-
pressor enters the bottom of an absorber, where the
remaining vinyl acetate is recovered. A liquid stream
from the base is recirculated through a cooler and fed
to the middle of the absorber. Liquid acetic acid that
has been cooled is fed into the top of the absorber to
provide the nal scrubbing. The liquid bottoms prod-
uct from the absorber combines with the liquid from
the separator as the feed stream to the distillation
column.
Part of the overhead gas exiting the absorber enters
the carbon dioxide removal system. This could be one
of the several standard industrial CO
`
removal pro-
cesses. Here we simplify this system by treating it as
a component separator with a certain eciency that is
a function of rate and composition. The gas stream
minus carbon dioxide is split, with part going to the
purge for removal of the inert ethane fromthe process.
The rest combines with the large recycle gas stream
and goes to the feed-euent heat exchanger. The fresh
ethylene feed stream is added. The gas recycle stream,
the fresh acetic acid feed, and the recycle liquid acetic
acid stream enter the vaporizer, where steam is used to
vaporize the liquid. The gas stream from the vapori-
zer is further heated to the desired reactor inlet tem-
perature in a trim heater using steam. Fresh oxygen is
added to the gas stream from the vaporizer just prior
to the reactor to keep the oxygen composition in the
gas recycle loop outside the explosivity region.
The azeotropic distillation column separates the
vinyl acetate and water from the unconverted acetic
acid. The overhead product is condensed with cooling
water and the liquid goes to a decanter, where the
vinyl acetate and water phases separate. The organic
and aqueous products are sent for further rening to
another distillation section. Here we ignore the addi-
tional separation steps required to produce vinyl acet-
ate of sucient purity because there is no recycle from
the rening train back to the reaction loop. The bot-
tom product from the distillation column contains
acetic acid, which recycles back to the vaporizer along
with fresh make-up acetic acid. Part of this bottoms
stream is the wash acid used in the absorber after
being cooled.
3. Physical property data
The vaporliquid equilibrium (VLE) data for the
three nonideal component pairs are given in Table 1.
These data come from the vapor-liquid equilibrium
data collection in the Chemistry data series published
by DECHEMA. VLE calculations are performed as-
suming an ideal vapor phase and a standard Wilson
liquid activity coecient model. This takes the form

GH
"

G
exp(!a
GH
/R), (5)
where is the absolute temperature in K, R is the gas
constant (1.987 cal/mol K), and
G
is the molar vol-
ume of component i given in DECHEMA and listed
in Table 1.
The Wilson parameters we use for the VAcH
`
O
pair are assumed to be the same as the parameters for
ethyl acetate and water. The reason for this assump-
tion is that no VLE data are presented in DECHEMA
for vinyl acetate and water, but ethyl acetate and vinyl
acetate are quite similar species and should behave
essentially identically. The liquid-liquid equilibrium
solubility data for the VAcH
`
O pair in the column
decanter come from Smith (1942) extrapolated to the
decanter temperature of 40C.
Acetic acid dimerizes in the vapor phase. The Wil-
son parameters listed in DECHEMA for the
H
`
OHAc pair assume the eect of dimerization is
modeled. Without considering the vapor-phase asso-
ciation, the DECHEMA parameters predict the exist-
ence of an azeotrope close to pure water. Such an
Study for the vinyl acetate monomer process 869
Table 2. Pure component physical properties (c
N
in cal/(g C))
Component Molecular Specic Latent Liquid Vapor
weight gravity heat ht capacity ht capacity
(cal/mol) ab ab
O
`
32 0.5 2300 0.30 0.2180.0001
CO
`
44.01 1.18 2429 0.60 0.230
C
`
H
"
28.05 0.57 1260 0.60 0.370.0007
C
`
H
'
30.05 0.57 1260 0.60 0.370.0007
VAc 86.09 0.85 8600 0.440.0011 0.290.0006
H
`
O 18.02 1 10684 0.990.0002 0.56!0.0016
HAc 60.05 0.98 5486 0.460.0012 0.520.0007
Table 1. Wilson parameters a
GH
and molar volumes
G
a
GH
VAc H
`
O HAc
G
(ml/mol)
VAc 0 1384.6 !136.1 93.1
H
`
O 2266.4 0 670.7 18.07
HAc 726.7 230.6 0 57.54
From DECHEMA vaporliquid equilibrium data collection
Vol. 1.
VAcH
`
O: Part 1b, p. 236.
VAcHAc: Part 5, p. 90.
H
`
OHAc: Part 1, p. 127.
azeotrope does not exist for this system. The VLE
behavior close to pure acetic acid is acceptable with-
out a model of dimerization. Since we operate in the
process where the VLE behavior is acceptable, we
have used the parameters in Table 1 without special
provisions for vapor-phase association.
Table 2 shows the pure component physical prop-
erty data, which we obtained from the DIPPR
database. These data include the molecular weight
M, the liquid specic gravity (based on the density
of water at 0C), the latent heat of vaporization H
T
at
0C (in cal/mol), and the liquid cJ
N
and vapor cT
N
heat
capacity parameters. The heat capacity expressions
we use have the following temperature dependence:
c
N
"a#bt, (6)
where c
N
is in cal/(gC) and t is the temperature inC.
Component vapor pressures PQ in psia (Table 3) are
calculated using the Antoine equation, with the Anto-
ine coecients listed in the DECHEMA volumes.
ln PQ"A#B/(t#C), (7)
where t is the temperature in C. For the four gas
components, the A parameters of the Antoine equa-
tion were estimated based upon the vapor pressure at
the operating conditions in the absorber. We removed
the temperature dependence to facilitate the dynamic
simulation. However, in the case of ethylene and
ethane, we found that we needed to include a small
Table 3. Component vapor pressure antoine coecients
!ln PQ"A#B/(t#C), where PQ in psia and t in C
Component A B C
O
`
9.2 0 273
CO
`
7.937 0 273
C
`
H
"
9.497 !313 273
C
`
H
'
9.497 !313 273
VAc 12.6564 !2984.45 226.66
H
`
O 14.6394 !3984.92 233.426
HAc 14.5236 !4457.83 258.45
temperature dependence for the bubble point calcu-
lations to function properly.
4. Process data and constraints
4.1. Design requirements
The process design that we use is based upon the
owsheet shown in SRI Report 15B. We assume that
the production basis of our process with new catalyst
is 785 mol/min VAc and at the given conditions
85 mol/min CO
`
is also produced. For a plant with
90% operating utility, this corresponds to an annual
production rate of 32;10' kg/yr, if the VAc rate is
sustained over the life of the catalyst. We assume that
the catalyst lifetime is one year.
The ethylene and oxygen feed streams come from
supply headers. Acetic acid comes from a storage
tank. The carbon dioxide is released to the atmo-
sphere. The gas purge stream is sent to a thermal
converter. The vinyl acetate and water products from
the decanter are fed to other distillation columns in
a rening train. Available on the plant are the follow-
ing utilities: cooling tower water at a supply temper-
ature of 30C, steam at supply pressures of 50 and
200 psia, refrigeration at !25C, and electricity and
process water. Economic data for raw material and
energy costs are listed in Table 4. Any capital equip-
ment and vessel cost data can be found in Guthrie
(1969). The costs should be updated to current prices.
Also, the appropriate material of construction factors
should be used. Cost correlations for some equipment
are given in Douglas (1988).
870 M.L. LUYBEN and B.D. TYRE US
Table 5. Process stream data Table I
Reactor Reactor Absorber Absorber Absorber Purge
in out vapor in vapor out liquid out ow
Stream Number 1 2 3 4 5 6
Flow (mol/min) 19250 18850 16240 15790 1210 3
Temperature (C) 148.5 158.9 80 40.4 47.7 40.4
Pressure (psia) 128 90 128 128 128 128
O
`
(mol frac) 0.075 0.049 0.057 0.058 0.001 0.059
CO
`
0.007 0.011 0.013 0.014 0.001 68
C
`
H
"
0.583 0.551 0.642 0.658 0.025 0.667
C
`
H
'
0.216 0.221 0.256 0.263 0.010 0.266
VAc 0 0.043 0.021 0.002 0.255 0.002
H
`
O 0.009 0.055 0.007 0.001 0.129 0.001
HAc 0.110 0.070 0.004 0.004 0.579 0.005
moles/million.
Pressure drop in gas loop assumed to be in reactor.
Table 6. Process stream data Table II
Column Column Organic Aqueous Fresh
feed bottoms product product HAc feed
Stream number 7 8 9 10 11
Flow (mol/min) 3820 2160 826 831 785
Temperature (C) 42.5 137.2 40 40 30
Pressure (psia) 84 30 18 18 150
VAc (mol frac) 0.206 11 0.950 0.002 0
H
`
O 0.281 0.093 0.050 0.998 0
HAc 0.513 0.907 370 370 1
moles/million.
Table 4. Economic data for vinyl acetate process
Item Cost/price
Acetic acid $0.596/kg
Oxygen $0.044/kg
Ethylene $0.442/kg
Vinyl acetate $0.971/kg
200 psia steam $11/1000 kg
50 psia steam $8.8/1000 kg
Cooling tower water $0.02/1000 l
Process water $0.15/1000 l
!25C refrigeration $0.12/h, ton
Electricity $0.065/kwh
Tables 57 contain the ow, temperature, pressure,
and composition data for selected streams in the pro-
cess. The corresponding stream numbers are shown in
Fig. 1. In our simulation, all gas is removed in a com-
ponent separator prior to the distillation column. This
involves the liquid from the separator and the absorb-
er. The gas is sent back and combines with the vapor
product from the separator to form the vapor feed to
the absorber. Tables 810 contain certain vessel data
that are required to size the equipment and construct
the simulation. These data come from our TMODS
dynamic simulation and not from a commercial
steady-state simulation package.
The reactor contains 622 tubes packed with cata-
lyst. The tube diameter is 3.7 cm and length 10 m.
Steam is generated on the shell side of the reactor to
remove the heat of reaction. We have modeled the
reactor in 10 sections in the axial direction. The reac-
tor temperature prole is shown in Fig. 2. The ow-
sheet design conditions are for a new catalyst with an
activity of 1. However, the catalyst does deactivate
over the course of operation. This deactivation via
sintering is a nonlinear function of operating time (t
WP
)
and temperature, since higher temperatures within the
tubes (t
RS@C
) promote deactivation. We assume that the
activity (a) decays exponentially with time from 1.0 to
0.8 after 1 yr according to
a"f (t
RS@C
) exp(!t
WP
/0.621). (8)
Study for the vinyl acetate monomer process 871
Table 7. Process stream data Table III
Fresh Fresh CO
`
CO
`
C
`
H
"
feed O
`
feed purge Removal
in ow
Stream number 12 13 14 15
Flow (mol/min) 831 521 85 6411
Temperature (C) 30 30 40.4 40.4
Pressure (psia) 150 150 128 128
O
`
(mol frac) 0 1 0 Same
CO
`
0 0 1 as
C
`
H
"
0.999 0 0 stream
C
`
H
'
0.001 0 0 4
Table 8. Reactor and vaporizer equipment data
Catalyst weight 2590 kg
Catalyst porosity 0.8
Catalyst bulk density 0.385 kg/l
Catalyst heat capacity 0.23 cal/g C
Overall heat transfer coecient 150 kcal/h C m`
Number of tubes 622
Tube length 10 m
Tube diameter 3.7 cm
Circumferential heat transfer area 725 m`
Shell side temperature 133C
Reactor heat duty 2.8;10' kcal/h
Steam drum volume 2 m`
BFW to steam drum 79.5 kg/min
Reactor feed heater duty 5.3;10` kcal/h
Vaporizer duty 1.3;10' kcal/h
Vaporizer total volume 17 m`
Vaporizer working level volume 4 m`
Vaporizer temperature 119C
If the tube temperature has not exceeded 180C,
then f (t
RS@C
)"1. Above this temperature, then
f (t
RS@C
)"exp[!(t
RS@C
!180)/50], where t
RS@C
is in C.
Two parameters dene the performance of the cata-
lyst: selectivity (SEL) and space-time yield (STY).
Catalyst selectivity determines the fraction of the
ethylene consumed that makes the desired vinyl acet-
ate product.
SEL"100
mol/min VAc
mol/min VAc#0.5 mol/min CO
`
.
(9)
For conditions at the design basis with fresh catalyst,
the selectivity is 94.8%. At a catalyst activity of 0.8,
higher reactor temperatures are required to achieve
about the same VAc production rate, increasing the
production rate of CO
`
to 126 mol/min and reducing
the selectivity to 92.4%. The spacetime yield quan-
ties the activity of the catalyst by volume.
STY"g VAc/h/liter catalyst. (10)
Table 9. FEHE, separator, and absorber equipment data
FEHE duty 4.4;10` kcal/h
FEHE hot outlet temperature 134C
FEHE A 6800 kcal/h C
Separator cooler duty 2.7;10' kcal/h
Separator volume 15 m`
Separator working level volume 8 m`
Gas loop volume 170 m`
Approximate compressor size 350 kW
Absorber base volume 8 m`
Absorber bottom section 2 theoretical stages
Absorber top section 6 theoretical stages
Absorber stage eciency 50%
Absorber tray holdup 14 kmol
Absorber liquid recirculation 15 kmol/min
Absorber cooler duty 6.5;10` kcal/h
Absorber wash acid feed 756 mol/min
Absorber wash acid cooler duty 1.3;10` kcal/h
Table 10. Column and decanter equipment data
Theoretical stages 20
Feed stage 15 from bottom
Stage eciency 50%
Tray holdup 2.3 kmol
Reboiler duty 4.0;10' kcal/h
Condenser duty 3.9;10' kcal/h
Base working level volume 6 m`
Decanter working level volume 5 m`
For conditions at the design basis, the STY is 603
since the total volume of catalyst (tube volume) is
6724 l.
The CO
`
removal system is assumed to be a com-
ponent separator that removes just carbon dioxide at
a certain eciency, which is the fraction in the feed
leaving in the CO
`
purge. This eciency (E) is
a function of the feed rate (F
''`
in mol/min) and
composition (x
''`
in mole fraction). At the design
conditions, the eciency is 0.995 for a feed rate
of 6410 mol/min at 0.014 mol fraction CO
`
. The
872 M.L. LUYBEN and B.D. TYRE US
Fig. 2. Reactor temperature prole.
maximum allowable feed rate to the CO
`
removal
systemis 8000 mol/min set by its capacity. The follow-
ing correlation determines the system eciency
E"0.995!3.14;10'(F
''`
!6410)
!32.5(x
''`
!0.014), (11)
where the eciency must lie between 0 and 1.
Two key safety constraints exist in the process.
First, the oxygen composition must not exceed 8 mol
% anywhere in the gas recycle loop to remain outside
the explosivity envelope of ethylene (Coward and
Jones, 1952). Continuous and reliable O
`
analyzers
will be installed in the process at the inlet of the
reactor to monitor oxygen composition. Second, the
pressure in the gas recycle loop and distillation col-
umn cannot exceed 140 psia because of the mechanical
construction limit of the process vessels. Pressure
measurements are readily available and will be instal-
led at appropriate locations. Exceeding either the oxy-
gen concentration or pressure limits will shut down
the process via interlocks.
Several other operational constraints must also be
satised during process operation. The peak reactor
temperature along the length of the tube must remain
below 200C, otherwise mechanical damage occurs to
the catalyst requiring shutdown. Liquid levels in the
vaporizer, separator, absorber base, distillation col-
umn base, and decanter must operate within the limits
of 1090%. The vessel volumes listed have been pro-
posed for the working liquid inventories between the
level taps.
Reactor inlet temperature must exceed 130C to
prevent condensation of liquid in the reactor. The hot
side exit temperature from the feed-euent heat ex-
changer (FEHE) must remain above 130C to avoid
condensation in the exchanger, which has been
designed to handle only vapor-phase ow. In the
azeotropic distillation column, the acetic acid in the
decanter organic phase must not exceed 600 mol/mil-
lion to prevent product contamination. A decanter
composition analysis for acetic acid is available from
the laboratory every 4 h. Also, the vinyl acetate com-
position in the bottoms stream must remain below
100 mol/million to minimize polymerization and foul-
ing in the column reboiler and vaporizer. The column
temperature prole is plotted in Fig. 3.
4.2. Control requirements
Figure 4 shows the location of the 26 control valves
that have been proposed in the preliminary design.
Flow, temperature, pressure, and level measurements
are readily available and can be installed in any loca-
tion needed for control. The O
`
analyzers at the
reactor inlet are specialized instruments that are fast
and reliable. However, if additional composition
measurements are used, they must be conventional
chromatographic analyzer types that are character-
ized by sampling frequency and reliability problems.
It must be assumed that any chromatographic ana-
lyzer used in this process has a 10 min sampling fre-
quency and 10 min deadtime. Also, this analyzer has
a 90% utility. The remainder of the time the instru-
ment is o-line for maintenance or calibration. Hence,
when an analyzer is not on line, the plant must be able
to continue producing vinyl acetate. Further, it must
be demonstrated how the control system functions
when an analyzer is down and one of the disturbances
listed below occurs.
The control system which we are asked to design
for this process must be able to operate in the face of
several known disturbances and changes we antici-
pate a priori.
1. The process will operate at a catalyst activity of 0.8
(or lower based upon reactor temperature) at the
end of 1 yr. The control system must still function
for the changed conditions.
2. Process operation must be regulated automatically
to reect changes in raw material and operating
costs so that the process always runs close to the
ecomomic optimum.
Fig. 3. Azeotropic distillation column prole.
Study for the vinyl acetate monomer process 873
Fig. 4. Location of control valves.
3. The control system must be able to change produc-
tion rate (as measured by steady organic ow from
the decanter) by at least 20% (both up and down)
over the course of 6 h. This is due to limits on tank
storage.
4. The plant needs to run at half the VAc production
design rate but at maximum selectivity. This is the
result of an occasionally precipitous drop in the
price of vinyl acetate to a third its normal value.
5. In-line (but not operating) spare pumps will be
installed for the fresh acetic acid supply line and for
the distillation column feed stream. However, it
must be demonstrated what the control system
does during the course of a 5 min loss of either
fresh acetic acid or column feed pump.
6. The control system must handle a step change in
the composition of ethane in the fresh ethylene feed
stream from 0.001 to 0.003 mol fraction.
7. The control system must not shut down the pro-
cess due to the loss of fresh oxygen feed ow.
Instead, this should result in the process going into
hot recycle mode.
5. Nonlinear dynamic model
5.1. Background
In this section we provide a brief description of the
nonlinear dynamic modules used in TMODS. These
modules have been constructed following the practi-
cal philosophy toward dynamic simulation outlined
in Luyben (1990). Our purpose is to provide a
feel for the level of detail we use in our dynamic
simulations without divulging exactly how our simu-
lator is implemented.
We would like to point out that the kind of dy-
namic models we use strike a balance between simula-
tion eciency and rigor of representation. We have
found that simulation speed is of vital importance and
we always strive to keep our user interactive simula-
tions running between 10 and 60 times real time.
There are basically two ways to meet this goal. We
can either limit the scope of the simulation to a few
process units that are rigorously simulated or limit the
complexity of most units within a plantwide scope.
Since we have found the plantwide perspective to be
most important for control simulations, we generally
opt for the second alternative.
This nonlinear dynamic model has been utilized to
conrm the feasibility of the simulation. It is the basis
for the data presented in the stream and equipment
tables. We also have used the simulation to test the
control strategy described in Luyben et al. (1997),
which was derived following the plantwide control
design procedure.
5.2. Physical property calculations
TMODS is implemented in an object-oriented
framework (Tyreus, 1992). It consists of a number of
generic classes that can be instantiated into objects in
the simulation. The objects are given appropriate
parameters to reect the actual piece of equipment
they represent. The objects are also connected to-
gether by the end user of the simulation to create
a owsheet. The most important aspect of the object-
oriented representation is the use of uid objects. The
874 M.L. LUYBEN and B.D. TYRE US
uid objects are instances of the uid system pertain-
ing to the simulation. The uid system contains all the
physical property constants described earlier includ-
ing the kinetic parameters. Each uid object is then
able to perform a number of services based upon these
parameters. For homogeneous (single phase) uids
these services amount to calculating thermodynamic
state variables from the knowledge of two intensive
state variables and the composition of the uid. In
TMODS these services are implemented to be explicit
functions of the known state variables. Examples are
"(P, , n

, n
`
,
2
),
h"h(P, , x

, x
`
,
2
),
"(P, u, x

, x
`
,
2
),
v"v(P, , x

, x
`
,
2
),
where P is the total pressure; is the temperature;
n
G
is the number of moles of component i; x
G
is the
mole fraction of component i (x
G
"n
G
/n
G
); is the
total internal energy; v is the specic volume, u is
the specic internal energy; and h is the specic
enthalpy.
In many cases the uid object represents a hetero-
geneous equilibrium system. Examples are the uid
objects in vaporizers, partial condensers, decanters,
and on the trays of a distillation column. Here the
thermodynamic state is completely specied by two
extensive state variables and the number of moles of
each component. The only dierence to a homogene-
ous system is that the unknown intensive state vari-
ables are no longer explicit functions of the known
entities. For example, to determine the pressure and
temperature of a uid in vaporliquid equilibrium, we
have to solve the following three nonlinear implicit
algebraic equations:
,

G
(K
G
!1)(n
G
/n)
1#(K
G
!1)
"0, (12)
/n!(1!) v*!v4"0, (13)
/n!(1!)u*!u4"0, (14)
where N is the number of components in the system,
n is the total number of moles in the uid object ( n
G
),
K
G
is the K-value for component i (K
G
"y
G
/x
G
), x
G
is the
mole fraction i in the liquid phase, y
G
is the mole
fraction i in the vapor phase, is the fraction of n in
the vapor phase; v is the specic volume; and u is the
specic internal energy.
The known state variables are total internal energy
and total volume that follow from the dynamic energy
balance and the equipment size. The total moles of
each component follow from the dynamic material
balances. The K-values are calculated from the equi-
librium requirement that the chemical potential of
each component is equal in both phases
4
G
"*
G
,
or equivalently that the partial fugacities of each com-
ponent are equal across the phases
f 4
G
"f *
G
.
In TMODS we assume that the vapor phase is ideal
such that
f 4
G
"y
G
P.
The liquid-phase partial fugacity is calculated with an
activity model according to
f *
G
"x
G

G
PQ
G
where
G
is the liquid-phase activity coecient of com-
ponent i and PQ
G
is the vapor pressure of component i.
To reduce the computational burden of iteratively
solving the three nonlinear equations, we frequently
make simplifying assumptions around two-phase sys-
tems. For example, in vaporliquid equilibria we of-
ten assume that the vapor holdup is negligible ("0).
This eliminates one of the three equations. The re-
maining variables can be solved for by making either
of the following assumptions. We can solve for and
P explicitly based on the fact that the n
G
s and the total
pertain to a single phase (the liquid). Or we can
assume that the pressure is known along with the n
G
s
of the liquid, and we can solve for temperature and
vapor compositions via a bubble point calculation.
The role of the unit operations is greatly simplied
by the use of uid objects. The unit operations contain
one uid object for each uid holdup in the equip-
ment. For example, a vaporizer contains one uid
object and a distillation tray section has a uid object
on each tray. The unit operation is responsible for
managing the accumulation of mass and energy into
the uid objects. With the knowledge of the total
internal energy, the total volume, and the number of
moles of each component, the thermodynamic state is
xed. The uid objects are then responsible for calcu-
lating all other relevant state variables pertaining to
the current state.
The equipment equations for accumulation of mass
and energy depend upon whether the system is lumped
or distributed.
5.3. Lumped equipment models
Material balances:
dn
G
dt
"FGLxGL
G
!FMSR xMSR
G
#R
G
. (15)
Energy balance:
d
dt
"FGLhGL!FMSRhMSR#Q#HR. (16)
Auxiliary relations:
FMSR"f (n
G
, P, , equipment conguration), (17)
where FGL is the molar ow of all streams entering the
vessel, xGL
G
is the mole fraction i in entering streams,
FMSR is the molar ow of exit streams, xMSR
G
is the mole
Study for the vinyl acetate monomer process 875
fraction i in exit streams, R
G
is the net production of
i from all chemical reactions, hGL is the specic en-
thalpy of entering streams, hMSR is the specic enthalpy
of exit streams, Q is the heat input to the vessel, and
HR is the total heat from reactions.
5.4. Distributed equipment models
Material balances:
*c
G
*t
"!
*J
G
*z
!
*(c
G
v)
*z
#
H

GH
r
H
!N
G
. (18)
Energy balance:
*(u
K
)
*t
"!
*J
O
*z
!
*(h
K
v)
*z
!
H
H
H
r
H
!
G
N
G
h
G
!q. (19)
Momentum balance:
*(v)
*t
"!
*P
*z
!
*(vv)
*z
#
G

G
FC
G
. (20)
Auxiliary relations:
N
G
"k
E
a(p
G
!p
*
G
), (21)
q"h
U
a(!
U
),
J
G
"!D
G
*c
G
*z
,
J
O
"!k
2
*
*z
,
where c
G
is the molar concentration of i; J
G
is the
diusion ux of component i; v is the bulk velocity of
uid; v
GH
is the stoichiometric coecient for compon-
ent i in reaction j; r
H
is the specic rate of reaction; N
G
is
the molar ux of component i; is the uid density;
u
K
is the specic internal energy (mass based); J
O
is the
heat ux due to conduction; h
K
is the specic enthalpy
(mass based); H
H
is the heat of reaction for reaction j;
q is the external heat ux per unit volume; FC
G
is the
external force acting on component i; k
E
is the overall
mass transfer coecient; a is the surface to volume
ratio for heat and mass transfer; p
G
is the partial
pressure for component i; p
*
G
is the interface partial
pressure for component i;
U
is the interface temper-
ature; D
G
is the molar diusivity coecient of compon-
ent i; and k
2
is the conductivity coecient.
5.5. Specic implementations
Catalytic plug ow reactor: The catalytic plug ow
reactor is modeled according to equations (18)(21)
on the tube side and equations (15) (17) on the shell
side. Simplifying assumptions are J
G
"0, J
O
"0, and
*(v)/*t"0. The time-independent momentum equa-
tion (20) sets the pressure prole in the reactor. We
assume that the entire gas loop pressure drop is repre-
sented by the reactor pressure drop. A simple back-
ward discretization scheme is used for the spatial
derivatives.
Gas Separator: This generic ash calculation can
be implemented in a number of dierent ways. In
TMODS this unit is implemented with two uid ob-
jects on the process side and a single-phase liquid
object on the shell side. In our implementation the
process side uid objects are not in equilibrium with
each other. The vapor object determines the system
pressure. The shell side uid determines the static
ash temperature. This allows us to use equation (12)
to solve for the partition of incoming feeds into the
vapor and liquid objects. The vapor object in the gas
separator denes the pressure level of the gas recycle
loop. As mentioned above, the reactor determines the
pressure drop in the loop.
Absorber: The gas absorber is implemented as two
countercurrent versions of equations (18), (19), and
(21). Each node, or stage, contains a liquid phase and
a vapor phase that are not in equilibrium with each
other. Instead, the single-phase state is determined by
the integration of equations (18) and (19). The auxili-
ary equations (21) then use the partial pressure and
temperature dierences between the two phases to
determine the mass and heat transfer rates.
aporizer: The vaporizer is implemented as a lum-
ped system with a single uid object. The vapor hold-
up is assumed negligible compared to the liquid
inventory.
Distillation Column: Each tray is a lumped system
and contains a uid object. The holdup in the vapor
phase is ignored. The pressure on each tray is assumed
known, which reduces the ash calculation to
a bubble point calculation. The energy balance deriv-
ative (16) is approximated numerically, which allows
us to solve for the vapor rate from stage to stage. This
is done to reduce system stiness.
Decanter: The TMODS decanter contains two
uid objects: one for the light phase and one for the
heavy phase. The partition coecients (K-values) are
assumed constant and independent of temperature.
This allows us to use equation (12) to determine the
distribution of the two liquid phases. Again, this is
done to save simulation time.
Heat Exchangers: Heat exchangers are calculated
statically with the eectiveness method. This allows
for an explicit calculation of the exit temperatures
based upon the exchanger eectiveness and the inlet
temperatures and heat capacities. The exchanger ef-
fectiveness depends on the eective A, the ratio
of stream heat capacities, and the exchanger cong-
uration. The exit temperatures are time-lagged to
introduce some realistic dynamics (usually very fast
compared to the overall recycle loop dynamics).
6. Conclusions
In this work we have presented design details of an
industrial process for the manufacture of vinyl acetate
876 M.L. LUYBEN and B.D. TYRE US
monomer. The design is preliminary and has not
been optimized. We have conveyed the study as if we
had been assigned the task of devising a control strat-
egy for this plant that is to be built. We have sum-
marized the design requirements for process opera-
tion and the control objectives that must be achieved
for various disturbances. A brief description was also
provided on the nonlinear dynamic modules in our
simulation. The purpose of this paper is to oer a real-
istic system that can be used by academic researchers
who are interested in working on an industrially rel-
evant study in the areas of design, simulation, and
control.
Complete models for this vinyl acetate process have
been made available by Aspen Technology and Hy-
protech. These models can be obtained electronically
from the following web sites:
(1) The application le for the vinyl acetate process
can be obtained from Aspen Technologys
example library Web site:
http://www.aspentec.com/tspsd/example/
example.htm
Search for Vinyl Acetate to nd and download
the application le.
(2) The case is made available at Hyprotech
www.hyprotech.com on the FTP site
(Papers/VinylAcetateProcess/VA

files.
zip). To download the le directly, users may
type the following into their Web browser:
ftp: //ftp.hyprotech.com/pub/Papers/
VinylAcetateProcess/VA

files.zip
Acknowledgements
We want to thank Dr W.D. Smith, Jr, DuPont, for
his support and help in making it possible to publish
this work. Also, we are grateful to Prof. W.L. Luyben,
Lehigh University, for his suggestions on the scope of
the paper and his careful review of the manuscript.
References
Coward, H.F. and Jones, G.W. (1952) Limits of ammability
of gases and vapors. Bulletin 503, Bureau of Mines.
Daniels, W.E. (1989) Vinyl ester polymers. In: Encyclopedia
of polymer science and engineering, 2nd ed. New York:
Wiley. Vol. 17, (pp. 393425).
Downs, J.J. and Vogel, E.F. (1993) A plant-wide industrial
process control problem. Computers in Chemical Engineer-
ing 17, 245255.
Douglas, J.M. (1988) Conceptual design of chemical processes.
New York: McGraw-Hill.
Guthrie, K.M. (1969) Capital cost estimating. Chemical En-
gineering 76, 114142. Mar. 24.
Luyben, M.L., Tyreus, B.D., and Luyben, W.L. (1997)
Plantwide control design procedure AIChE J. 43,
31613174.
Luyben, W.L. (1990) Process modeling, simulation and control
for chemical engineers, 2nd ed. New York: McGraw-Hill.
Nakamura, S. and Yasui, T. (1970) The mechanism of the
palladium-catalyzed synthesis of vinyl acetate from ethy-
lene in a heterogeneous gas reaction. Journal of Catalysis
17, 366374.
Neurock, M., Provine, W.D., Dixon, D.A., Coulston, G.W.,
Lerou, J.J. and van Santen, R.A. (1996). First principle analy-
sis of the catalytic reaction pathways in the synthesis of vinyl
acetate. Chemical Engineering Science 51, 16911699.
Samanos, B., Boutry, P. and Montarnal, R. (1971) The mech-
anism of vinyl acetate formation by gas-phase catalytic
ethylene acetoxidation. Journal of Catalysis 23, 1930.
Smith, J.C. (1942) The solubility diagrams for the systems
ethylidene diacetateacetic acidwater and vinyl acet-
ateacetonewater. Journal of Physical Chemistry 46,
229232.
Tyreus, B.D. (1992) Object-oriented simulation. In: W.L.
Luyben, (Ed.), Practical distillation control. chap. 5. New
York: Van Nostrand Reinhold.
Study for the vinyl acetate monomer process 877